Clustering-based Human Locomotion Parameters for Motion Type Classification

The paper proposes a classification method of human locomotion types from video sequences based on motion parameters clustering. A set of motion parameters is semi-automatically extracted from training video sequences that contain three different types of movement: walking, jogging and running. The motion parameters (postural, frequential, and cinematic) are stored in a relational database and statistic parameters such as minimum, maximum, average and standard deviation are computed. Then a K-means clustering is applied on all the statistic parameters combinations and the results are evaluated using the purity measure to determine most significant parameters to be used in classification. Because the number of training video sequences is reduced, the proposed method may be used as a model of classification only. The automatic determination of movement parameters will increase the data collection size and real testing of the classification method.